Differential Expression of Lipid Metabolism-Related Proteins in Different Breast Cancer Subtypes

Purpose This study aimed to determine the expression and clinical significance of proteins that are involved in lipid metabolism in human breast tumors. Methods Tumors from 476 breast cancer patients were used to construct tissue microarrays. Then, immunohistochemistry (IHC) for hormone-sensitive lipase (HSL), Perilipin 1 (PLIN1), fatty acid binding protein 4 (FABP4), carnitine palmitoyltransferase IA (CPT-1A), acyl-CoA oxidase 1 (ACOX-1), and fatty acid synthase (FASN) was performed on these microarrays. Results Breast tumors were classified into 4 subtypes: luminal A (n = 242; 50.8%), luminal B (n = 134; 28.2%), human epidermal growth factor receptor 2 (HER2) (n = 50; 10.5%), and triple negative breast cancer (TNBC) (n = 50; 10.5%). The expression of PLIN1 (p < 0.001), FABP4 (p = 0.029), CPT-1A (p = 0.001), ACOX-1 (p < 0.001), and FASN (p < 0.001) differed significantly among these tumor subtypes. Notably, PLIN1, CPT-1A, and FASN expression was highest in HER2 tumors and lowest in TNBC tumors. Similarly, the expression of FABP4 and ACOX-1 was highest in HER2 tumors and lowest in luminal A tumors. In addition, ACOX-1 positivity was associated with significantly shorter overall survival (p = 0.018). When tumor subtype was considered, FABP4 positivity was associated with significantly shorter disease-free survival (p = 0.005) and overall survival (p = 0.041) in TNBC. Conclusion Lipid metabolism-related proteins are differentially expressed in different IHC subtypes of breast cancer and some are associated with decreased survival rates.


Introduction
In oncology, the Warburg effect describes a significant metabolic change in energy production from oxidative phosphorylation in normal cells to aerobic glycolysis in cancer cells [1]. However, this definition is somewhat simplistic as tumors can use several different metabolic mechanisms to produce energy, depending on the type of tumor [2], which complicates targeted delivery of metabolic inhibitors to cancer cells. One such mechanism is lipid metabolism, which involves lipid synthesis, lipid degradation and catabolism, and fatty acid (FA) oxidation. Lipid synthesis includes FA and triacylglycerol (TAG) synthesis by fatty acid synthase (FASN) [3], and lipid degradation and catabolism include TAG, cholesterol ester, and phospholipid hydrolysis. A key enzyme in this process is hormone-sensitive lipase (HSL) [4][5][6]. FA oxidation involves catabolism of free fatty acids in the mitochondria to produce energy. Carnitine palmitoyltransferase IA (CPT-1A) and acyl-CoA oxidase 1 (ACOX-1) are two important enzymes in this process [5][6][7]. Aside from this process, lipid transport and uptake are indeed an important and under-appreciated aspect of lipid metabolism in cancer [8,9]. Two important proteins in this process are fatty acid binding protein 4 (FABP4), which transports free fatty acids, and Perilipin 1 (PLIN1), which helps to regulate triacylglycerol storage by suppressing its hydrolysis [10]. In addition, it is also becoming clear that lipid droplets are more than just passive storage components and are important in cancer as well, in particular for survival under stressful conditions [11,12], where lipid droplet proteins (HSL and PLIN1) play an important role.
Due to the clinical, histological, and molecular heterogeneity of breast tumors, many classification schemes have been proposed to group tumors with similar features. For example, gene profiling analyses of breast tumors have suggested 5 molecular subtypes, namely, luminal A, luminal B, HER2, normal breast-like, and basal-like) [13]. Since these subtypes differ in terms of their histology, clinical behavior, and therapeutic response, it is not surprising that they use metabolic pathways differentially. Indeed, previous studies have shown that proteins that are involved in glycolysis [14,15], glutaminolysis [16], and glycine or serine metabolism [17] are differentially expressed among different tumor subtypes. However, little is known about the differential expression of proteins that are involved in lipid metabolism in different breast cancer subtypes. As a result, this study aimed to determine the expression and clinical significance of proteins that are involved in lipolysis and mitochondrial β-oxidation in different breast cancer subtypes. protein from each sample was separated by SDS polyacrylamide gel electrophoresis and then blotted onto nitrocellulose membranes (Bio-Rad). The membranes were blocked with 7% nonfat dry milk in phosphate-buffered saline with Tween 20, and then incubated with primary antibodies against HSL, PLIN1, FABP4, CPT-1A, ACOX-1, and β-actin (Table 1) for 1 hour at room temperature. Then, the membranes were incubated with a horseradish peroxidaseconjugated secondary antibody for 1 hour at room temperature. Finally, the bound antibodies were visualized by using an enhanced chemiluminescent reagent (GE Healthcare Life Sciences). All antibodies were purchased from Abcam.

Patients selection
The study group consisted of 476 patients who were diagnosed with invasive breast cancer and underwent surgical excision at Yonsei University Severance Hospital (Seoul, Korea) between January 2002 and December 2006. Patients who received preoperative hormonal therapy or neoadjuvant chemotherapy were excluded from the study. This study was approved by the Institutional Review Board of Yonsei University Severance Hospital.
Hematoxylin and eosin (H&E)-stained tumor sections were retrospectively reviewed by a breast pathologist who graded the tumor using the Nottingham grading system [18]. For each patient, age at the time of initial diagnosis, lymph node metastasis, tumor recurrence, distant metastasis, and survival were recorded.

Construction of tissue microarrays
A representative area was selected from a H&E section and the corresponding area was marked on the surface of the paraffin-embedded tissue block. Then, a paraffin tissue punch was used to extract a 3 mm core sample from the selected area, which was placed into a 6 × 5 recipient block. Two tissue cores were extracted to reduce sampling bias. Each core was assigned a unique tissue microarray location number, which was linked to a database that contained other clinicopathological data.

Immunohistochemistry
All immunohistochemical staining was performed on formalin-fixed, paraffin-embedded tissue sections. Briefly, 5 μm-thick sections were cut with a microtome, transferred onto adhesive slides, and then dried at 62°C for 30 minutes. All slides were incubated with primary antibodies (Table 1). After applying primary antibodies, blocking time was 2 hours at 37°C. Subsequently, immunodetection was performed by using a commercial streptavidin-biotin kit according to the manufacturer's instructions, which involved incubation with biotinylated anti-mouse or anti-rabbit immunoglobulin, followed by peroxidase-labeled streptavidin and 3,3 0 -diaminobenzidine chromogenic substrate. The primary antibody incubation step was omitted in the negative control. Finally, the slides were counterstained with Harris hematoxylin.

Interpretation of immunohistochemical staining
The status of all immunohistochemical markers was determined by using light microscopy to assess the fraction of stained cells. HSL, PLIN1, FABP4, CPT-1A, ACOX-1, and fatty acid synthase (FASN) immunostaining were scored as the product of the proportion of stained cells (0 = no staining, 1 = less than 30%, or 2 = more than 30%) and staining intensity (0 = no staining, 1 = weak, 2 = moderate, or 3 = strong). The scores for the proportion of stained cells and the staining were multiplied to provide a total score. A total score of 2-6 was considered positive, while a score of 0 or 1 was considered negative [19]. Similarly, Ki-67 immunostains were scored as the percentage of stained tumor cells, which was defined as the Ki-67 labeling index (LI). In addition, estrogen receptor (ER) and progesterone receptor (PR) positivity were defined as one percent or more cells having positively stained nuclei [20].
HER2 immunohistochemistry (IHC) results were classified according to American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines, which includes the following categories: 0 = no immunostaining; 1+ = weak/incomplete membrane staining in less than 10% of tumor cells; 2+ = complete membrane staining that is either uniform or weak in at least 10% of all tumor cells; and 3+ = uniform, intense membrane staining in at least 30% of tumor cells. HER2 positivity was defined as IHC 3+, while IHC 0 or 1+ were considered to be HER2 negative [21]. However, IHC 2+ is an equivocal classification, so in these cases, HER2 expression was further evaluated with fluorescence in situ hybridization (FISH).
Fluorescence in situ hybridization (FISH) analysis FISH was performed by using a PathVysion HER2 DNA Probe Kit (Vysis, Downers Grove, IL, USA) according to the manufacturer's instructions. Then, the HER2 gene copy number was quantified by using an epifluorescence microscope (Olympus, Tokyo, Japan). At least 60 tumor cell nuclei from three separate regions were used to measure signals from DNA probes specific for HER2 and the centromeric region of chromosome 17 (CEP17). HER2 gene amplification was determined according to ASCO/CAP guidelines [21]. Specifically, HER2 negativity was defined as a dual-probe HER2/CEP17 signal ratio less than 1.8 with an absolute HER2 gene copy number of less than 4 signals per cell. On the other hand, a HER2/CEP17 ratio greater than 2.2 with an absolute HER2 gene copy number of greater than 6 signals per cell was considered to be HER2 positive. HER2/CEP17 ratios between 1.8 and 2.2 or absolute HER2 copy numbers between 4 and 6 were considered to be HER2 equivocal.

Statistical analyses
Student's t test and Fisher's exact test were used to detect statistically significant differences (p < 0.05) in continuous and categorical variables, respectively. Kaplan-Meier survival curves and log-rank statistics were also used to estimate disease-free survival (DFS) and overall survival (OS) rates. Multivariate regression analysis was performed with the Cox proportional hazards model. All statistical analyses were calculated with SPSS for Windows, version 12.0 (SPSS Inc., Chicago, IL, USA).

Patient characteristics
The clinicopathological characteristics of the 476 female patients in this study are shown in   (Table 3 and Fig. 2).

Association of the expression of lipid metabolism-related proteins with prognosis
The only statistically significant relationship revealed by univariate analysis was an association between ACOX-1 positivity and shorter OS (p = 0.018; Table 5). However, when tumor subtype was considered, there were also significant associations between FABP4 positivity and shorter DFS (p = 0.005) and shorter OS (p = 0.041) in TNBC (Fig. 3). Multivariate Cox regression analysis revealed that higher T stage is an independent predictive factor for shorter DFS (hazard ratio: 3.801, 95% CI: 1.361-10.61, p = 0.011), but no other parameter had a statistically significant association with either DFS or OS ( Table 6).
The effect of the expression of lipid metabolism-related proteins on survival rates depending on ER, PR, and HER2 status of breast tumors is shown in Fig. 4. ACOX-1 positivity was associated with shorter OS in patients with ER positive tumors (p = 0.024) and HER2 positive tumors (p = 0.023). In addition, FASN negativity was associated with shorter DFS in patients with PR negative tumors (p = 0.046).

Discussion
In this study, we observed differential expression of lipid metabolism-related proteins in 4 breast cancer subtypes. In particular, HER2 tumors showed the highest expression of PLIN1, CPT-1A, FASN, FABP4, and ACOX-1. This finding is consistent with two previous studies about HER2 and FASN. In one study, FASN expression was strongly correlated with HER2 status [23]. Another study showed that HER2 interacts with FASN and promotes FASN phosphorylation, which increases its activity and leads to cancer cell proliferation, and eventually, metastasis [24]. Our results suggest that lipid metabolism in breast tumors with the HER2 subtype is higher than that in other subtypes. Previous studies showed the difference of lipid metabolism genes according to breast cancer subtypes: products of de novo fatty acid synthesis, such as palmitate-containing phosphatidylcholine, were high in ER negative and grade 3 breast cancers [25], and acyl-CoA:cholesterol acyltransferase 1 (ACAT) activity was high in ER-negative basal-like breast cancer [26]. However, there was one report that the gene for secreted phospholipase A2 (sPLA2) is silenced in triple negative breast cancer cells [27]. Therefore, further study is required to evaluate the status of lipid metabolism genes according to breast cancer subtypes. More generally, our results support the hypothesis that different breast cancer subtypes use different metabolic pathways, which has been previously suggested [2]. In this study, the expression of three lipid metabolism-related proteins (PLIN1, CPT-1A, and FASN) in TNBC cells was lower than that in other subtypes. Although a previous study demonstrated that the gene for secreted phospholipase A2 (sPLA2) is silenced in triple negative breast cancer cells [27], another report showed that sPLA2-induced lipid droplet formation plays an important role in TNBC cell proliferation and survival during starvation [12], indicating controversial results. Specifically, TNBC is often associated with a high histologic grade, accelerated mitosis, tumor necrosis, aggressive clinical behavior, and poor prognosis [28,29]. In addition, glycolysis-related proteins are highly expressed in TNBC and basal-like breast tumors, which imply high glycolytic activity [14,15], suggesting high metabolic activity in TNBC. Therefore, further study of lipid metabolism activity in TNBC is required.
The preference of different breast cancer subtypes for different metabolic pathways may have clinical implications. For example, ACOX-1 positivity was significantly correlated with shorter OS as well as clinicopathological characteristics, such as higher histologic grade (p = 0.006), ER negativity (p < 0.001), PR negativity (p < 0.001), and higher Ki-67 LI (p < 0.001), that are associated with poor prognosis in breast cancer patients. These results are in accordance with a previous study that showed that ACOX-1 expression increases with worsening histologic grade in brain glioma [30]. Similarly, FABP4 positivity was significantly associated with shorter DFS and OS in TNBC patients. These findings are supported by the observations that serum FABP levels are significantly higher in breast cancer patients than in healthy patients, and high serum levels of FABP are associated with adverse tumor characteristics, such as large tumor size and lymph node metastasis [31]. Similar correlations have also been found in bladder cancer [32] and prostate cancer [33].   These differences suggest that selective inhibition of lipid metabolism-related proteins may be a potential chemotherapeutic strategy for some breast cancer subtypes. This hypothesis is supported by the observation that inhibitors of glycolysis-related molecules, such as hypoxiainducible factor 1α [34,35], glucose transporter 1 [36,37], carbonic anhydrase IX [38], and monocarboxylate transporter 4 [39], suppress tumor progression in several types of cancer, including breast cancer. In addition, several FASN inhibitors have been shown to decrease tumor cell growth or increase tumor cell death [24,40,41]. However, further studies are needed to determine the value of lipid metabolism-related proteins as therapeutic targets.   In conclusion, our results showed that lipid metabolism-related proteins are differentially expressed in different breast cancer subtypes, which may aid the development of novel chemotherapeutic agents.

Author Contributions
Conceived and designed the experiments: SK JSK. Performed the experiments: SK YKL. Analyzed the data: SK YKL JSK. Contributed reagents/materials/analysis tools: JSK. Wrote the paper: SK JSK.